Project description
Global assessments of land carbon and water fluxes based on machine learning models trained on eddy covariance flux tower observations have been providing an important complementary perspective to assessments by process-models. These data-driven products are particularly strong at extracting and generalizing the dominant signals of observed land-atmosphere flux variations such as the diurnal and seasonal cycles. However, smaller but very important signals like anomalies due to extreme conditions, or trends are uncertain largely due to the limited information content in the spatially clumped and often sparse flux tower observations. The objective of this PhD is to address this problem by incorporating process-knowledge into the machine learning framework in the form of a hybrid model. We expect that incorporating theoretical constraints can alleviate lacking data constraints. Like process-models such hybrid models also represent internal diagnostic variables that can facilitate a much better interpretability and attribution of derived flux patterns.
Working group
The successful candidate will work in the Department of Biogeochemical Integration at the Max-Planck-Institute for Biogeochemistry. The working group provides long-standing expertise and experience of the different fields relevant to that project: Machine learning (Christian Reimers, Alexander Brenning, Markus Reichstein, Martin Jung), carbon cycle processes (Markus Reichstein, Martin Jung, Jacob Nelson), hybrid modelling (Markus Reichstein, Martin Jung, Christian Reimers), and flux tower measurements (Jacob Nelson, Markus Reichstein, Martin Jung). The candidate will benefit from and closely collaborate with the FLUXCOM team at MPI-BGC on the overall methodological approach. Further collaborations with members of the Michael-Stifel-Center Jena for Data-driven and simulation science as well as within the European Lab for Learning and Intelligent Systems (ELLIS) are likely to emerge.
Requirements for the PhD project are
Applications are open to highly motivated and independent students from any country who have:
- A Master's degree in Computer Science, Physics, Environmental Sciences or a related discipline
- programming experience, preferably in Python
- background in machine learning, preferably neural networks
- basic understanding of land-atmosphere carbon and water fluxes
- Very good English written and communication skills.
The Max Planck Society (MPS) strives for gender equality and diversity. The MPS aims to increase the proportion of women in areas where they are underrepresented. Women are therefore explicitly encouraged to apply. We welcome applications from all fields. The Max Planck Society has set itself the goal of employing more severely disabled people. Applications from severely disabled persons are expressly encouraged.
|